Use of Mobile EEG in Decoding Hand Movement Speed and Position

With the majority of brain–computer interface (BCI) research currently restricted to the controlled settings in labs, there is a growing interest to study the feasibility of BCI applications in the real world. This interest is largely driven by the availability of ergonomic electroencephalography (EEG) recording devices. In this article, we investigate the feasibility of applying a commercial EEG device in motor decoding and propose a signal processing strategy to classify and reconstruct hand movement speed and position. An experiment is designed to simultaneously record EEG and hand position, while the user executes movement toward the left or right direction at two different speeds. A visual interface is designed to guide the user to perform the task under each condition. Data are recorded from 21 subjects. The classification of direction-dependent and -independent speeds is implemented using spatial and spectral features. An optimized movement parameter estimation strategy is proposed to reconstruct the instantaneous position and speed of the hand from EEG. Average performances of $73.36\ ({ \pm \! 11.95})\% $ , $69.46\ ({ \pm \! 13.39})\% $ , and $68.98\ ({ \pm \! 14.79})\% $ are obtained for direction independent, right and left direction fast versus slow classification, respectively. The average correlation between recorded and reconstructed hand position and speed along the x-axis is in the range of [0.22, 0.57] and [0.26, 0.58], respectively. The reported results validate the use of commercial-grade EEG and low-frequency components of EEG relevant to motor decoding for real-world motor kinematics BCI applications.